1、Update NVIDIA driver
Choose the driver that corresponds to your graphics card (choose the studio version, not the game version).
2, add Anaconda Tsinghua mirror
Method 1: anaconda command replacement
conda config --add channels /anaconda/cloud/conda-forge conda config --add channels /anaconda/cloud/msys2 conda config --add channels /anaconda/cloud/pytorch/ conda config --add channels /anaconda/pkgs/r conda config --add channels /anaconda/pkgs/free conda config --add channels /anaconda/pkgs/main conda config --set show_channel_urls yes
(Mark) Switch back to the default source code:
conda config --remove-key channels
Method 2: Replace .condarc
show_channel_urls: true channel_alias: /anaconda default_channels: - /anaconda/pkgs/main - /anaconda/pkgs/free - /anaconda/pkgs/r - /anaconda/pkgs/pro - /anaconda/pkgs/msys2 - /anaconda/cloud/pytorch/ custom_channels: conda-forge: /anaconda/cloud msys2: /anaconda/cloud bioconda: /anaconda/cloud menpo: /anaconda/cloud pytorch: /anaconda/cloud simpleitk: /anaconda/cloud
3. Create a virtual environment
establish： conda create -n environmental name python= opens： activate environmental name cloture： conda deactivate removing： conda remove -n environmental name --all Add Packages： conda install -n environmental name package name Removal Packages： conda remove -n environmental name package name
4. Installation of pytorch commands - corresponds to the situation of choice: link to the command code (pytorch official website)
Installation (recommended)
pip install torch===1.4.0 torchvision===0.5.0 -f /whl/torch_stable.html
This one's more stable than the conda, and hitting enter when it gets stuck seems to save the day.
Installation (not recommended, always interrupted)
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch (To remove-c pytorch Otherwise, it's still the default source.) The final input command: conda install pytorch torchvision cudatoolkit=10.1
The network's down all over the place. Sometimes it's down again at 48%.
Test after installation
import torch
flag = .is_available()
print(flag)
ngpu= 1
# Decide which device we want to run on
device = ("cuda:0" if (.is_available() and ngpu > 0) else "cpu")
print(device)
print(.get_device_name(0))
print((3,3).cuda())
Result: after being tossed around by the conda command all afternoon, I was finally rescued by the pip command!
summarize
To this Anaconda configuration pytorch-gpu virtual environment steps to organize the article is introduced to this, more related to Anaconda pytorch-gpu virtual environment content, please search for my previous articles or continue to browse the following related articles I hope you will support me in the future more!